The officialization of romaji
« previous post | next post »
Japan might finally switch to romaji system you already use
Japan Today (June 30, 2025)
I've read many articles of this sort, but I cite this one because it is fairly recent and is from a reputable newspaper.
If you’ve spent any time learning Japanese or just getting around Japan, you’ve probably come across romaji — the Roman alphabet version of Japanese. It shows up on signs, maps, train stations and in most textbooks for foreign learners. But not all romaji is the same. Depending on where you look, you might see shi spelled as shi, si or even something else.
Romaji (ローマ字) means “Roman letters,” using the Latin alphabet to write Japanese sounds. It’s not a full writing system like hiragana or kanji, but it’s a useful bridge for learners and travelers who can’t yet read Japanese….
Japan has more than one romaji system — and now, for the first time in 70 years, the government is considering switching its official system to something more globally friendly.
Romanization is widely used in Japan, not least for entering text in digital devices. What makes it seem not so ubiquitous is the fact that there are competing systems, one government sanctioned, and one used popularly by a large proportion of the public.
Japanese has a few romaji systems; the two big ones are Kunrei-shiki and Hepburn.
The Hepburn system was developed in 1867 by American missionary James Curtis Hepburn. His goal was to help English speakers pronounce Japanese more accurately, so Hepburn spells kana like し, ち, and つ as “shi,” “chi” and “tsu,” matching their actual sounds.
Japan officially adopted the Kunrei-shiki system in 1937 as the government’s standard. Kunrei-shiki spells Japanese sounds based on their arrangement in the kana alphabet rather than how they sound to English speakers.
For example, it writes し as “si,” ち as “ti,” and つ as “tu” because they belong to the “s” and “t” groups in the kana chart, even though they sound like “shi,” “chi” and “tsu.”
However, after World War II, the American occupation under General Douglas MacArthur preferred the Modified Hepburn system, seeing Kunrei-shiki as linked to Japan’s militaristic past. It was probably also way easier to read for him. Despite Kunrei-shiki remaining the official system taught in schools, Hepburn became the more common style used in daily life and public signage.
While the Kunrei-shiki structure is logical for native speakers and linguists, it can be confusing for anyone relying on English pronunciation. If you saw “tikatetu (subway)” on a sign, would you read it as “chikatetsu”? Probably not.
Though Kunrei-shiki (Cabinet-order style) has been the official, government-mandated romanization system in Japan since 1937, frequently used in school education, official documents, and by the National Diet Library. However, it is largely being replaced by the widely used Hepburn romanization system for international purposes, travel, and public signage. (AIO)
Why Japan Might Make Hepburn Official
While the Kunrei-shiki system remains the official government standard and is still taught in schools for grammar, it’s not the version most people encounter daily.
Instead, Hepburn is widely used in places designed for readability and especially where international communication is important, like train stations, maps and tourist spots. Even outside of Japanese schools, most native Japanese people use the Hepburn system. That gap between policy and everyday practice is why Japan’s Agency for Cultural Affairs is reviewing the rules.
In a recent draft proposal, a government panel recommended officially switching to Hepburn as the standard romanization system because the Kunrei-shiki has never really caught on with the general public, while Hepburn has become the de facto norm.
This would make spellings like “Tokyo” (instead of “Toukyou”) the standard because that’s how most people actually write and read Japanese these days. This should help clear up confusion and make getting around Japan easier for everyone.
If the government goes ahead with the switch, you’ll start seeing familiar Hepburn spellings become the official standard on everything from textbooks to prefectural names. In truth, foreigners may not notice a change at all—when was the last time someone texted you to meet at “Sinzyuku station” instead of Shinjuku station?
Something you might find shocking, that occurs smack dab at the conclusion of the introductory "History" section of the major Wikipedia article on the "Romanization of Japanese", is a brief account of rōmaji "As a replacement for the Japanese writing system" that has this striking sentence:
Today, the use of Nihon-shiki for writing Japanese is advocated by the Oomoto sect and some independent organizations.
What is the Oomoto sect?
Oomoto (大本, Ōmoto; lit. "Great Source" or "Great Origin"), also known as Oomoto-kyo (大本教, Ōmoto-kyō), is a religion founded in the 1890s by Deguchi Nao (1836–1918) and Deguchi Onisaburō (1871–1948). Oomoto is typically categorized as a Shinto-based Japanese new religion. The spiritual leaders of the movement have always been women within the Deguchi family….
As we have been learning during the last few weeks and months, women have been at the forefront of phonetic writing, not just in Japan, but in China and Korea as well.
All Japanese who have attended elementary school since World War II have been taught to read and write romanized Japanese. Therefore, almost all Japanese can read and write Japanese by using rōmaji.
Selected readings
- "What good are kanji?" (10/21/25)
- "Women's script wins in the end" (1/19/26)
- "Women's writing: dead or alive" (10/2/20)
- "Women's Romanization for Hong Kong" (8/17/19)
- "Japanese Romanization: they still haven't decided" (10/2/22)
- "Japanese Romanization: they still haven't decided, part 2" (11/8/22)
- "Rōmaji dialog between 'bread' and 'tea''" (8/9/15)
- "Differential retention of sinographs across East Asia" (12/12/25)
- "'Romanisation 'gives clarity'" (5/19/23)
- "How to learn to read and write Chinese" (8/13/19)
- "How to learn Mandarin" (7/17/18)
- "How not to learn Chinese" (4/16/17)
- "How to learn Chinese and Japanese" (2/17/14)
JMGN said,
February 1, 2026 @ 8:14 am
Has the issue been solved with KAYII and HAYEE?
https://ibb.co/x8wJtD0T
[url=https://postimg.cc/WhYPdkbX][img]https://i.postimg.cc/WhYPdkbX/KAYII-AND-HAYEE.jpg[/img][/url]
Philip Taylor said,
February 1, 2026 @ 8:35 am
I would have ascribed more credence to the article had it been published in a Japanese-language newspaper …
Jonathan Smith said,
February 1, 2026 @ 1:29 pm
"Shi" "chi" and "tsu" to "[match] their actual sounds".
Chris Button said,
February 1, 2026 @ 2:00 pm
@ JMGN
I'm not sure what you are referring to specifically. You could also throw "wu" in alongside "yi", but what do you need them for (I don't think their kana are even sanctioned since the distinctions didn't really exist)
As for "ye", it is indeed very confusing how Ebisu (from Webisu) is written as Yebisu.
Jim Breen said,
February 1, 2026 @ 6:02 pm
Unfortunately, that Japan Today article has a number of errors, the most glaring of which is that the "Hepburn system was developed in 1867 by American missionary James Curtis Hepburn". Hepburn didn't develop the system named after him; it got that label because he used it in the 2nd edition of his dictionary. The system was developed within the 羅馬字会 (Romanization Club), of which Hepburn was a member.
murawaki said,
February 1, 2026 @ 6:51 pm
We can distinguish between two types of romanization:
1. a system designed to represent a language's sound system as a whole, and
2. a system used mainly to embed words and phrases from that language into another language's text.
For Korean, McCune-Reischauer is clearly of type (2). South Korea's Revised Romanization strikes me as an unsatisfying halfway compromise. It ends up satisfying neither purpose well.
I see the national adoption of the Hepburn system as a farewell to the attempt to make type (1) work for Japanese.
Victor Mair said,
February 1, 2026 @ 8:29 pm
Well put, murawaki!
Neil Kubler said,
February 1, 2026 @ 8:54 pm
There is no "perfect" romanization system for everyone; it all depends on who and for what purpose. As Murawaki pointed out, Kunreisiki is better for representing the Japanese sound system from the point of view of native speakers (and remember some of those who helped create it hoped it would eventually replace kanzi and kana). Kunreisiki is also better for serious Japanese learners as it greatly simplifies the description of Japanese grammar: cf. matu/matte/matimasu (all with the same verb stem mat-) vs. Hepburn matsu/matte/machimasu (with 3 different verb stems). Sure, Hepburn is better for helping English speakers approximate the pronunciation of Japanese, but it won't help speakers of other languages nearly as much.
Victor Mair said,
February 1, 2026 @ 9:10 pm
Superbly and elegantly expressed, Neil!
Mike Ryan said,
February 2, 2026 @ 4:22 am
Neil Kubler said,…(all with the same verb stem mat-) …
Is not the stem of 待つ "ま" (ma) ? （ま）つ、（ま）って、（ま）ちます
Michael Watts said,
February 2, 2026 @ 5:12 am
I'm curious here about the difference between "native speakers" and "people who have been educated in traditional kana tables". It happens that those tables position つ tsu in a row with other syllables that start with /t/ (and don't continue into /s/). Literate Japanese all know this and therefore the belief exists that tsu is in some sense part of a family with ta, te, to, and of course chi.
But it's not obvious to me that this is the natural position of "native speakers". Do they find it difficult to pronounce /tɯ/, or is it more that there are no words that feature it? How much do Japanese students struggle to pronounce the English word "to"?
The kana tables also tell us that chi exists (in the T row), and that cha, chu, and cho exist (in the separate CH row), but that che does not and cannot exist. Here, I feel more confident: the position of a native speaker independent of traditional education is definitely that chi belongs in the CH row with cha, chu, and cho, not in the T row with ta, te, and to. And I suspect that che is an unproblematic syllable for native speakers to pronounce. I see that the Japanese wikipedia page for Che Guevara names him チェ, which appears to be exactly what you'd expect to find in the CH row if "che" had a place in it at all… but it doesn't.
If everyone in Japan forgot their kana tables, and a group of native speakers developed a new syllabary from scratch, what are the odds that it would reflect the odder parts of kunrei-shiki?
Michael Watts said,
February 2, 2026 @ 5:18 am
Well, there is a link to a Japanese source. (The link is broken, but the article was archived: here )
That source is in English, but the publishing organization offers this self-description:
We are the international service of NHK, the sole public media organization of Japan. Through television, radio and internet we reach the entire world in multiple languages, providing accurate and impartial news and rich, quality content.
That said, NHK's reporting says "the panel plans to complete its recommendation by around this autumn [2025]." Presumably, there is more recent news on the matter.
Peter Cyrus said,
February 2, 2026 @ 7:31 am
Can we label Kunreishiki as "phonemic" and Hepburn as "phonetic"? Neither is 100%, but the main difference is the level of transcription.
Musa writes Japanese at an even lower level, which I call allophonic, for instance writing ga as ŋa when it's pronounced that way. I suspect that isn't done more often only because we don't have a letter for ŋ in the English alphabet. Likewise, we write the alternance between affricates and fricatives, and even voiceless vowels.
But from my point of view, the biggest flaw in both romanizations is the absence of marking the pitch accent.
Jonathan Smith said,
February 2, 2026 @ 8:12 am
@Michael Watts I'm pretty sure (correct me) ち つ truly involve conditioning before -i -u i.e. the ability to say (e.g. English) "tea" with a phonetic dental stop, etc., is (relatively speaking) sophisticated~newfangled. So "T row" is English thinking; should be "た row". (Of course large-scale borrowing can break such systems over time…)
And then the likes of "cha" etc. are written with "ち" not for no reason.
J.W. Brewer said,
February 2, 2026 @ 9:49 am
I was struck by the article's sentence 'This would make spellings like “Tokyo” (instead of “Toukyou”) the standard because that’s how most people actually write and read Japanese these days.' Leaving aside whether "Toukyou" is the technical Kunrei-shiki answer (sources differ …), the *proper* Hepburn would be "Tōkyō." This is a perfect illustration of the curse of many romanization systems – diacriticals are meaningful but frequently get omitted in practice. In Europe, folks like Croats and Poles manage to use their own diacriticals pretty consistently, but I'm not sure that romaji-using Japanese are there yet and how they could get there.
To be clear, "Tokyo" is IMHO the standard and correct English spelling of the city's name. But Tōkyō is the spelling in romaji-scripted Japanese.
Victor Mair said,
February 2, 2026 @ 11:34 am
And how about "romaji" vs. "rōmaji"?
J.W. Brewer said,
February 2, 2026 @ 6:26 pm
@VHM: well, I was writing in English and "romaji" is an English word with an English spelling (and an English pronunciation, in which the GOAT vowel is not stretched into a second mora). That might not be the best transliteration of ローマ字, though.
Michael Watts said,
February 2, 2026 @ 7:27 pm
I don't think that actually contradicts what I said about ち. Stipulate that [ti] is difficult to say.
It is nevertheless the case that Japanese has a /t/ phoneme and, separately, a /tɕ/ phoneme (I'm just taking wikipedia's word for the conventional IPA transcription), and that ち is seen as beginning with /tɕ/. The traditional kana table implies that ち shares its onset with た, but (a) this is not true, and (b) native speakers do not perceive it as true.
(A good test of this would be to find some Japanese alliterative verse where ち is alliterated.)
Imagine two rows of a hypothetical full table: ta, ti, tu, te, to / cha, chi, chu, che, cho.
Japanese education teaches that "chi" and "che" do not exist. All 5 "t-" syllables do, but ti and tu are pronounced differently than they are spelled.
I am saying that, in a table developed from scratch by native speakers of the modern language, "chi" would exist and be assigned to the same category as "cha", "chu", and "cho". Under the assumption that [ti] cannot be realized, a system that reflected modern (instead of ancient) phonology would imply that "ti", not "chi", didn't exist.
Joseph Tomei said,
February 2, 2026 @ 8:09 pm
I briefly looked back at other posts, so apologies if this has been covered, but when you talk about halfway compromises, Japan is second to none.
The fight between Hepburn and Kunrei was so bitter and inconclusive that the government, to try and split the baby, had Hepburn for 'outward facing' uses (i.e. when foreigners (largely english speakers) might use/read it) while Kunrei was for 'inward use'. So transport signage (romaji for Foreigners) was in Hepburn, but some government documents and other uses (romaji for internal use) was in Kunrei. This was a particular problem when Japanese started getting credit cards, because Shoji Chiba on a passport would be Syozi Tiba and be rejected.
Japanese banks are still very insistent that the name on the account be the same as the name on the document, and now, with increased scrutiny because of terrorism, a number of foreigners are having problems because they sometimes have their middle name, the order is reversed or the romanization is not exactly the same. (fun article, unfortunately behind a paywall, about German spouses, because of the design of their passport, getting geb (short for geborene/ the spouse's maiden name) stuck in their name. (This facebook group suggests that the new design of the German passport may solve this problem, but it is amazing that it requires the Germans to redesign their passport rather than the Japanese looking at what the geb means)
https://www.facebook.com/wegmitgeb/
Kunrei shiki was also taught to elementary school students (because it was to help with learning the logic of Japanese) but Hepburn was taught to JHS students. When I first came to Japan 30+ years ago, it was still not unusual to encounter students who had learned Kunrei shiki too well, and found it handicapping their English. With the move (mistaken in imo) of teaching English in elementary school, Hepburn has now been adopted as the de facto standard (the transition is only being finalized this year, which I think accounts for the slew of articles hailing Hepburn adoption)
On a separate point, about Oomoto Kyo was the religion connected to Ueshiba Morihei, the founder of aikido, and a martial art that, along with Judo, left behind the iemoto system of transmission (where knowledge is passed through a single lineage ) so I'm wondering if that outward impulse that romanization indicates is of a piece with that.
Joseph Tomei said,
February 2, 2026 @ 8:58 pm
Sorry, left off the link to the geb article
https://www.japantimes.co.jp/community/2014/09/08/issues/mind-geb-little-word-big-problem-japans-german-residents/
I'll leave any other mistakes in my post as a monument to bad eyesight and poor proof reading.
Michael Watts said,
February 3, 2026 @ 1:56 am
What is the concept behind "romaji for internal use"? Why not just use written Japanese? Shouldn't Shoji Chiba's name be represented in kanji as far as his credit card and bank are concerned?
Joseph Tomei said,
February 3, 2026 @ 4:43 am
For the credit cards, when they were only stamped pieces of plastic, the letters were embossed and that allowed the business to put a special form that had a carbon paper in it, allowing it to create an exact copy of the letters for a physical record of the transaction. Kanji, because of both complexity and variety, would be impossible to emboss. However, kunrei shiki kana still appears in all my bank transactions that appear in a bank book or printed receipt, so I'm thinking that despite the heralded shift, those will stay the same.
Internal use, the most prominent was to teach children from the 4th grade on how to represent Japanese in roman letters. It was also used for things like library records, maps that were 'for' Japanese and database entry. The division between 'internal' and 'external' breaks down pretty quickly, but pre war and post war when very few people could leave the country, it works after a fashion, if only to placate the two sides.
Marshall Unger discusses the debate and the ferocity of the sides in his book Literacy and Script Reform in Occupation Japan: Reading Between the Lines. I don't totally agree with all his points, but the history he discusses and how the Japanese ended up with a dual romanization system that is only now being phased out is pretty interesting.
Jarek Weckwerth said,
February 3, 2026 @ 8:52 am
"matching their actual sounds" — This, and some of the discussion posts, is a sad reminder of the detrimental effects of the hegemony of English. The "outward facing" system assumes that English grapheme-to-phoneme associations are the default.
"ch" and "sh" do not have an "actual sound".
One manifestation of this is the representations of East Slavic names in passports etc. These days, they have become standard in Poland and other Slavic countries due to their official, often obscuring the pronunciation of the originals for the uninitiated, despite the existence of local romanizations of Cyrillic. Kherson and surzhyk. FFS (excuse me).
Sorry, just a pet peeve of mine.
Michael Watts said,
February 3, 2026 @ 6:44 pm
But this comment / complaint remains coherent if you restrict yourself entirely to Japanese and pretend that English doesn't exist. As such, I don't see how it can be "a sad reminder of the detrimental effects of the hegemony of English".
In the world without English, it's still true that learning to write the first syllables of 茶色い, 中国, 調印, and 地位 as tya, tyu, tyo, and ti draws a distinction that doesn't exist and raises questions about why one of those onsets fails to match the others. If the official spellings were "f1", "f3", "f5", and "m2", there would be no question of resembling English, but the question of "matching the actual sounds" would be just as real.
Similarly, in a world with only English, you might see a push to respell colonel "to better match its actual sounds", and this would make sense even though colonel is already an English word. It's an English word with what can only be described as an ateji spelling, and this creates mismatch issues that are not related to whether English is a better language than other languages.
Josh R. said,
February 3, 2026 @ 7:11 pm
Michael Watts said:
"Literate Japanese all know this and therefore the belief exists that tsu is in some sense part of a family with ta, te, to, and of course chi."
That belief exists because tsu is part of a phonemic family with ta, te, to, and chi, not because people learn the kana tables at school. Verbs that end with "tsu" conjugate regularly through the ta line (e.g, matsu, "to wait", matanai, machimasu, matsu, mate, matou).
Likewise, verbs that end with "su" regularly conjugate through the sa line, including the irregular "shi".
Naturally, these conjugations long predate the particular order of the current kana tables, which, while existing since antiquity, only came into widespread use in the 20th century. (Previously, non-scholars utilized the "Iroha" ordering, based on a poem which utilized every kana once.)
Joseph Tomei said,
February 3, 2026 @ 7:18 pm
To reflect on the hegemony of English (which starts with the hegemony of the roman script), it is probably not helpful to look at this at a granular level, but to think of it in this way.
A person goes to do something that requires a form. They go in and on the form, they are asked to write their name. They write it in their native language (kanji, Cyrillic, etc) and the clerk says sorry, that's not your name, you need to write your name "in English" (i.e. Roman letters). The person says that's not my name, my name is this, pointing to the script on the paper. At this point, if the person and the clerk were equivalent in power, they would be able to discuss why this is and eventually they would decide which to use. But that doesn't happen.
If you argue about data and transferability, all of this happened well before the advent of computers and the internet. In fact, it is similar to the questions that Europe is having now about having 'digital sovereignty' and is a good way to consider what kind of trade-offs are going to happen by considering what has happened in terms of language in these places.
The discussion about internal use vs external use (my clumsy wording, perhaps others have better naming) was precisely about this. The kunrei-shiki folks wanted their language to be represented by the romanization that they chose, not a romanization that they felt was imposed on them. That the kunrei shiki backers lost (after roughly a century of debate) precisely points to the hegemony of English and why people argue that it is inevitable.
Jonathan Smith said,
February 3, 2026 @ 7:54 pm
It's not that simple: I guess Japanese is in some intermediate state where [tɕi] has /t-/, as was historically the case, for certain speakers/contexts, but rather /tɕ-/ for other speakers/contexts. E.g. verb conjugation supports the first intuition (complementary point to Neil Kubler's) whereas e.g. knowledge of lots of English loans showing /t/ ≠ /tɕ/ supports the second. You are just insisting on the second analysis — which maybe does more reflect a contemporary Japanese intuition, IDK.
Re: Kunrei tya tyû tyô vs. ti, same sorts of issues exist in any phonological-ish spelling system. Here all syllables begin with the "t" and there is if you like a spelling rule "ty" > "ti" in the absence of a second vowel symbol. Kinda makes sense right? Whereas Hepburn cha chū chō chi use… the letter "h" as in shi (!) and in ha (!?). Problematic right?
Since I look at it every day, I am personally frustrated by Taiwanese Church Romanization "ch" vs. (aspirated) "chh" — why not "c" vs. "ch" and save boatloads of ink right? So it is.
anon said,
February 3, 2026 @ 9:21 pm
Some linguistic documentations remind me that Japanese [t͡s], [t͡ʃ], [ɸ] are allophones of /t/ and /h or p/.
anon said,
February 3, 2026 @ 9:56 pm
…[tʃ] could be allophone of /s/. And the status of /ʃ/ as in shi, sho, shu may be not phonemic as well.
Recent katakana transcriptions of loan words may help the phonemization of certain phonemes.
Michael Watts said,
February 3, 2026 @ 11:01 pm
OK? It's not like those are equally valid arguments. One of them is a phonological argument, and the other one… isn't.
There is nothing to suggest that conjugational endings can't take whatever arbitrary form they feel like. There is nothing to explain about a conjugational paradigm that alternates different phonemes in a certain position. I understand that Gaelic noun inflection marks the accusative case by mutating whatever follows the word that is in the accusative, meaning the change occurs to an arbitrary onset of a different word. Should we conclude that Gaelic cannot exist?
Jonathan Smith said,
February 3, 2026 @ 11:18 pm
? I said "support… intuition". Try this if you prefer: e.g. [t] and [tɕ] don't contrast in native vocabulary, plus they alternate… soooooo some native intuition in favor of /t/.
Why so resistant to the idea of this question being not 100% straightforward though?
Jarek Weckwerth said,
February 4, 2026 @ 12:48 pm
@ Joseph Tomei: Beautifully put, thank you!
@ Michael Watts: I'm not arguing about contrasts. I'm saying that claiming that "ch" is the evident solution for those sounds is anglocentric. In French, "ch" is associated with /ʃ/, in Italian, with /k/, and in German, with /x/. It does not have an "actual sound".
Philip Taylor said,
February 4, 2026 @ 1:01 pm
« I'm saying that claiming that "ch" is the evident solution for those sounds is anglocentric. In French, "ch" is associated with /ʃ/, in Italian, with /k/, and in German, with /x/. It does not have an "actual sound" » — hear hear. But even "anglocentric" doesn't really go far enough — in British English we have (for example) "cheese", "chiasma", "loch", "parachute", … so « "ch" does not have an 'actual sound' in many languages, including British English ».
Michael Watts said,
February 4, 2026 @ 6:27 pm
But this is false. [t] and [tɕ] do contrast in native vocabulary. There is no contrast between [ti] and [tɕi], but the contrast between /t/ and /tɕ/ is robust. Compare 関東 かんとう with 浣腸 かんちょう.
Jonathan Smith said,
February 4, 2026 @ 10:08 pm
^ That's (historically) Chinese, if it matters… anyway!
Re: general topic of this thread, question is how problematic is it that Kunrei has no (ready?) means of distinguishing [tɕ] from [t] etc. like in Hepburn and now in kana — and OK more broadly whether it's still really true that "Kunreisiki is better for representing the Japanese sound system from the point of view of native speakers" as Neil Kubler / others have it.